import numpy as np
import pandas as pd
import os
import matplotlib.pyplot as plt
import matplotlib
import seaborn as sns

import skimage.io as io
from skimage import util

from glob import glob
import trackpy

import ast
from ipywidgets import interact, widgets


# import code from the src
### import code from src folder
import sys
sys.path.append('../src/')
import chow_test_v
import area_analysis

import warnings
warnings.filterwarnings('ignore')

from sklearn.metrics import r2_score, median_absolute_error, mean_absolute_error
from sklearn.metrics import median_absolute_error, mean_squared_error, mean_squared_log_error

Read data

file_list=glob("../data/*_Dark_prepared.tif")
images_dict={}
for file_name in file_list:
    images_prepared=io.imread(file_name)
    file_prefix=os.path.split(file_name)[1].split('.')[0].split("_prepared")[0]
    
    binary_file=file_name.split(".tif")[0]+'_segmented'+'.tif'
    images_binary=io.imread(binary_file)
    tracks=pd.read_csv("../data/tracks_overlay_id_"+file_prefix+'.csv')
    tracks_big=pd.read_csv("../data/tracks_overlay_id_"+file_prefix+'_big.csv')
    img_dict={"img":images_prepared,"binary":images_binary,"tracks":tracks,"tracks_big":tracks_big}
    
    images_dict[file_prefix]=img_dict

Area

• The objects have dynamic behavior. The area measurements are fluctuate with a time. My try to divide objects on subunits does not work (geometry method, local-maximum method) .
• Area are generally grow in two ways: linear growth of the object itself (linearly?) and aggregation.

Area_replicas=pd.DataFrame()

for file_prefix in images_dict.keys():
    tracks=images_dict[file_prefix]["tracks_big"]
    images=images_dict[file_prefix]["img"]
    binary=images_dict[file_prefix]["binary"]

    # set up grid for plt plot
    sns.set(font_scale=2)
    Ngroups=tracks.groupby('particle').ngroups
    Nrows=np.int(np.ceil(Ngroups/5))
    fig,ax=plt.subplots(nrows=Nrows,ncols=5,figsize=(40,Nrows*7.5))
    axx=ax.ravel()

    
    #analyse each trajectory separately
    # 1) anomaly detection
    # 2) chow test anomalies
    # 3) fit according anomalies detection
    counter=0
    Area_file_prefix=pd.DataFrame()
    
    for id, traj in tracks.groupby('particle'):
        df_result=area_analysis.area_analysis(traj,axx[counter])
        Area_file_prefix=pd.concat([Area_file_prefix,df_result])
        
        if df_result.shape[0]==2:
            handles, labels = axx[counter].get_legend_handles_labels()
        counter=counter+1
        
    fig.suptitle(file_prefix,x=0.5, y=0.80)
    fig.legend(handles[:-1], labels[:-1],ncol=len(labels),loc='lower left',bbox_to_anchor=(0.1,0.90),fontsize="large")

              
    Area_file_prefix["file_prefix"]=file_prefix
    fig.savefig("../results/Area_updated_fit__{}.png".format(file_prefix))
    
    Area_replicas=pd.concat([Area_replicas,Area_file_prefix])
    

Barplots

Area

Area_replicas=Area_replicas[Area_replicas['len']>15]

sns.set(font_scale=1)
ax = sns.boxplot(x="slope", y="file_prefix",data=Area_replicas, showfliers = False)
ax.set_ylabel("Experiment");
ax.set_title("LinearRegression Slope");

quantile_25=lambda x: np.quantile(x, 0.25)
quantile_75=lambda x: np.quantile(x, 0.75)
table= pd.pivot_table(Area_replicas, values='slope',index=['file_prefix'], aggfunc={'median',quantile_25,quantile_75}) 
table=table.rename(columns={'<lambda_0>':'25% quantile','<lambda_1>':'75% quantile' })
table=table[["median","25% quantile","75% quantile"]]
table

	median	25% quantile	75% quantile
file_prefix
43121_Dark	1.276692	2.227256	0.743985
43122_Dark	0.797398	1.263010	0.419337
43123_Dark	0.627298	1.166823	0.412558
43125_Dark	1.188831	2.194574	0.620694

MAE

sns.set(font_scale=1)
ax = sns.boxplot(x="MAE", y="file_prefix",data=Area_replicas, showfliers = False)
ax.set_ylabel("Experiment");
ax.set_title("LinearRegression MAE");
ax.set_xlim([0,35]);

quantile_25=lambda x: np.quantile(x, 0.25)
quantile_75=lambda x: np.quantile(x, 0.75)
table= pd.pivot_table(Area_replicas, values='MAE',index=['file_prefix'], aggfunc={'median',quantile_25,quantile_75}) 
table=table.rename(columns={'<lambda_0>':'25% quantile','<lambda_1>':'75% quantile' })
table=table[["median","25% quantile","75% quantile"]]
table

	median	25% quantile	75% quantile
file_prefix
43121_Dark	9.648571	12.891673	7.279192
43122_Dark	11.854498	17.658272	7.895317
43123_Dark	13.524531	17.456520	8.727378
43125_Dark	13.549622	18.063856	10.815029

sns.set(font_scale=1)
ax = sns.scatterplot(x="slope",y="MAE",hue="file_prefix",style="file_prefix",data=Area_replicas)

ax.set_xlim([-15,15]);

Length

sns.set(font_scale=1)
ax = sns.boxplot(x="len", y="file_prefix",data=Area_replicas, showfliers = False)
ax.set_ylabel("Experiment");
ax.set_title("Length");

quantile_25=lambda x: np.quantile(x, 0.25)
quantile_75=lambda x: np.quantile(x, 0.75)
table= pd.pivot_table(Area_replicas, values='len',index=['file_prefix'], aggfunc={'median',quantile_25,quantile_75}) 
table=table.rename(columns={'<lambda_0>':'25% quantile','<lambda_1>':'75% quantile' })
table=table[["median","25% quantile","75% quantile"]]
table

	median	25% quantile	75% quantile
file_prefix
43121_Dark	20.0	20.00	20.00
43122_Dark	49.0	36.75	49.00
43123_Dark	83.0	48.25	97.00
43125_Dark	61.0	37.50	92.25

sns.set(font_scale=1)
ax = sns.scatterplot(x="len",y="MAE",hue="file_prefix",style="file_prefix",data=Area_replicas)

Intersept

sns.set(font_scale=1)
ax = sns.boxplot(x="intersept", y="file_prefix",data=Area_replicas, showfliers = False)
ax.set_ylabel("Experiment");
ax.set_xlim([0,3500]);
ax.set_title("Length");

sns.set(font_scale=1)
ax = sns.scatterplot(y="slope",x="intersept",hue="file_prefix",style="file_prefix",data=Area_replicas)
ax.set_title("Does the glownth rate depend on the initial size of the colony?");
ax.set(xscale="log");

PairPlot

Area_replicas=Area_replicas[Area_replicas['slope']<30]
select=['slope', 'intersept','MAE','len','file_prefix']
sns.pairplot(data=Area_replicas[select],hue='file_prefix');